Notes from working concepts out in code — networking, distributed systems, algorithms and observability. Mostly build journals from side projects, written while the details were still fresh.
Crawling a Cloudflare-protected PurePortal site into MongoDB, building a per-field TF-IDF index from scratch, and ranking with cosine similarity — plus what the evaluation numbers actually exposed about a bag-of-words engine.
How I turned a hand-written Vite single-pager into a content engine: Astro 7 content collections, build-time Pagefind search, and a deploy pipeline on Cloudflare Workers that fires on every git push.
Fitting five polynomial regression architectures to four EEG input channels using nothing but normal equations, then comparing them with RSS, log-likelihood, AIC and BIC, validating on a held-out split, and running rejection ABC — plus the two places where the numbers quietly disagreed with the method.
Notes on moving tree layout out of the client and into stored geometry, then rendering only the viewport across three levels of detail — so draw cost tracks what's on screen instead of how big the tree is.
Working through NAT traversal and bulk transfer over an unreliable datagram protocol — rendezvous, simultaneous open, keepalives, chunking, integrity and flow control — by building a peer-to-peer file sharing system in Go.
Working out the forward pass, softmax with cross-entropy, the chain rule and mini-batch gradient descent by building a multi-layer perceptron in plain Java — no ML libraries — and learning why data layout decides what the algorithm can express.
Working through structured logging, trace correlation, agent-based instrumentation and the ship-parse-store-visualise pipeline by wiring a Spring Boot service to Elasticsearch, Logstash, Kibana and Elastic APM in Docker Compose.
Working through constraint search — pruned tree exploration, candidate checks, recursive descent and undo — by building a browser Sudoku solver that animates its own search using async/await to yield to the event loop.
Working through genome design, fitness shaping, selection, mutation and elitism by building a browser simulation where a population of dots evolves its way to a goal — and learning what population diversity actually buys you.
No articles in this category yet.